JPH0273477A - Method and device for intensifying digitalized image - Google Patents

Abstract

PURPOSE: To easily perform emphasis (modification) for the multi-level picture elements of digitized images at a low cost by using a picture element window to be scanned, that is moved, provided with many picture elements and provided with a center picture element. CONSTITUTION: The multi-level picture element of the digitized image are supplied to the input 6 of a window shift register (3×3 window) 16 and a shift register 12. Then, the output of the register 12 is inputted to the input 8 of the window 16 and the shift register 14 and the output 13 from the register 14 is inputted to the window 16. Then, respective picture element values inside the window 16 are shifted respectively by clock signals, the output is inputted through an annular intermediate value circuit part 18 and a central deviation circuit part 20 to circuits 22 and 24 and the emphasized (modified) picture element D'c of the center picture element Dc , of the window 16 is obtained. Then, the picture element D'c is compared with a threshold value from a memory 111 in a comparator circuit 12, a mark is generated by an imaging device or the like by using the picture element D'c when it is larger or equal and it is not generated when it is smaller. Thus, the digitized image is easily emphasized at a low cost.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of the Invention 1] The invention disclosed herein relates to a method and apparatus for enhancing a digitized image in which multi-level pixels of a digitized image are enhanced by non-uniform mask processing, and the center pixel Background on the value of the center pixel that is modified by modulating the value of the center pixel using a moving window, which is composed of a number of pixels including a center pixel, if desired, when the value is reduced to 1 bit by a threshold value. The value is determined and used.

BACKGROUND OF THE INVENTION As a direct application of minicomputers and word processors, many computer output devices have been developed that accept digital data and output hard copies. Currently, most of these devices use non-impact printing methods, such as ink jet or laser printing. These machines are faster and quieter than impact printers, yet have sufficient resolution for letter quality printing.

In related technology, charge-coupled devices (CODs)
) As the resolution of image scanning arrays decreases, current forms of analog photocopying machines digitize documents by scanning, dividing the image on the document into pixels from multiple pixel regions, and for each pixel After obtaining a digital value at one of a number of levels, it is possible to develop a more versatile machine that prints using digital data. Furthermore, digital data can be input into a computer, and the computer can directly power the printing machine. This is just one example of new technology using digital data.

By using a general-purpose computer, the digital enhancement method can be applied to 25 levels, although not limited to 256 levels.
Using six levels to modify the density of each image element or pixel can be applied to improve the perceived quality of images generated from digital signals. This has not been commercially realized because the computational process to enhance the document is too time-consuming or expensive;
It is necessary to simplify the algorithm used to reduce this computational load. digital documents are
Since 15g1 of pixels would be required to determine this, it would take several minutes to several hours to enhance the document with a general-purpose computer. Specialized image enhancement computers can also be used, but the cost can range from tens of thousands of dollars to hundreds of thousands of dollars.

In any image forming device, there are usually two formats:
That is, (2) deterioration of the original information, which is represented by blurred contours (lower Wt image resolution, lower sharpness) and random irregularities (noise, fog), occurs. In a typical photographic image, the original information is obscured by a mask of moderate negative contrast (usually lower than that of the original).

) has been known for some time that it is possible to enhance the sharpness of contours and reduce noise by masking them. An early invention by J. A. C. Yule described this photographic masking method (U.S. Patent No. 2.4
07.21 November; No. 2.420.636: No. 2.45
No. 5,849), and a more complex method is described in U.S. Patent No. 3.
No. 615.433.

Early attempts to use raster scanning to image while photographically measuring the instantaneous light intensity and attenuating the beam according to a predetermined relationship with the light intensity
olse) in U.S. Pat. No. 3,011,395. The rapid evolution of the space program has led to the emergence of highly efficient digital means to digitize, analyze, reconstruct, and enhance images. Median filtering has also been investigated as a means of enhancing contour contrast (e.g. Frieden
JO8A 66.280~283 (19
76)).

In the field of medical radiography, this generally promoted the development of computerized tomography (0 machines) and digital processing of radiographs (AMD (S,
R, Astey) et al. 5PrE 314.327
~330 (1979)) and Wilson (C, 1,
Wilson et al. gsPLE 314.327-33
0 (1981)). In these solutions, the image was divided into a large number of pixels by scanning.

A moving window of n× pixels centered at pixel i with image faD is examined by an on-line computer as it scans through pixel i.

The average value of the pixels within the window, ie the arithmetic mean mo, is used to qualify the center pixel ID, to the value filtered by the following algorithm: /.

DI -aD, -bU The parameters a and b are chosen to give a special image character, but are constant for one scan.

The idea of variable parameters a and b throughout the scan of the image based on certain local characteristics of the image has been investigated and published in several patents (If
, Kato et al., Yoneda Patent No. 4゜315.31
No. 8 and No. 4,317.179, Ishida (H, l5h
U.S. Patent No. 4.346.409 by ida) et al.)
discloses a specific relationship between the parameters and the values of the corners or D that can further enhance the image. However, these techniques do not distinguish between noise and image contours as far as enhancement is concerned; the greater the center pixel value, or D, the greater the enhancement. Similar solutions have been used in other technical fields. Therefore, Albarslau (E,^1parslau) and Indje ([.

InCQ) Paper by Ic (TEEE Vol, SM
C.

376-384.1981), images are processed by a contour enhancement algorithm based in part on adaptive parameters determined by the difference between the maximum and minimum pixel values in a window of arbitrary points.

No. 4,334,244 processes the running average of a group of digitized samples and the median value of a group of digitized samples of a video signal to obtain slope samples. This slope sample is evaluated according to the nozzle level of the video signal. The estimated slope samples are multiplied by the difference between the intermediate mean sample and the running mean sample and added to the mean sample to generate an output sample. The output samples are converted to an enhanced video signal. It is also known to modify analog photographic images. These can be expanded or combined, and have contrast that emphasizes some areas and suppresses others. The same results can be obtained for these by digitally processing the digital images. Furthermore,
Digital image processing can modify images in many important ways that analog processing cannot. Edge enhancement, noise filtering, gray swirl stretching or compression, and color correction are just a few of the techniques available to digitally improve the perception of images.

Images with accented contours can increase visual depth and, in some cases, often appear subjectively more desirable because they provide more information. Since sharpness is a contour-related effect, a method is needed to amplify contours independently. The contour represents the high frequency components of the image. That is, areas where the intensity of the image changes rapidly have high spatial frequencies. Conversely, regions where the image changes slowly have only low spatial frequencies.

A known method for extracting high frequency components from a signal is to remove the low frequency components from the original signal.

By appropriately combining the all-bus filter and the low-pass filter as shown in the following equation, a synthesized high-pass filter can be created.

High pass filter = (all pass filter) - (low pass filter > (1, 1) When made independent, by multiplying this filter by the coefficient (k),
The synthesized high frequency components can be evaluated.

k x high pass filter = kx (all pass filter) - kx (low pass filter > (1, 2) The synthesized filter removes the low frequency components from the original image and converts it into image data. and needs to be added to this image data, resulting in an unsharp masking enhancement filter (USM) that can be expressed as:

USM - (kx all pass filter) - × (low pass filter) 10 (low pass filter) (1, 3) or USM - kX (all pass filter) - (k - 1
) (Low Pass Filter') (1, 4> Iteratively processing the image data through the filter of the method described above results in enhanced image data.

The above unsharp masking enhancement filters are described by Mahmoodi (A, B., Hahmoodi) and Nelson (
Owen L, Nelson) for US Patent 4,
It is used in No. 571゜635,
A sliding window is used that moves across the image pixel signal both vertically and horizontally to input the image pixel signal into the filter. The level or value of the center pixel (D,) of the window is processed as the value of the all-pass filter, and the middle level, that is, the middle 1ffi (1
! (D) and represents the unsharp masking enhancement filter. Using ', Equations 1 and 4 can be expressed as follows.

D′ −kDc −(k−1)D (2,1)
4.571.6 for US patent by Mahmoudi et al.
This is the format after formulas (2, 4) of No. 35 are corrected. The error in formula (2, 4) in the above US patent is due to the error described in line 511m, line 30 of the specification of that patent, and in order to obtain formula (2, 4) of that patent, This is what is used. In line 30, b/1-b-
1-k is described, but this is -b/1-b-1-
It should be.

By transforming (2,1), it can be expressed as the following equation.

D , =k(D −D>+D (2,2>C 4,571°635@
makes the coefficient k adaptable to changes in the image scene. If the contours intersect, it is desirable to enhance the image contours and contrast (with a larger factor k).

In areas without contours, keeping the enhancement factor at a low level minimizes high spatial frequency noise and smooths the image. The algorithm of this patent controls an emphasis factor k with a statistical parameter that can be measured for any given window location. The standard deviation (ρ) of the pixel values within the window is used to generate the appropriate function f(ρ) for determining k.

The low cost and high speed implementation of the algorithm disclosed in U.S. Pat. , and the lack of specialized integrated circuits currently available. The lack of specialized integrated circuits capable of spatial image processing as described in U.S. Pat. No. 4,571.635 means that
This means that complex image processing algorithms must be implemented using a large number of low-level circuit elements. This results in the use of circular algorithms and requires many circuit boards with large power supplies and complex interconnections, as described in US Pat. No. 4.571.
The complexity and cost of implementing the algorithms disclosed in No. 635 are significant in applications such as television receivers, microfilm readers/printers, copy machines, facsimile machines, optical character readers, and digital still photo cameras. This results in a cost/benefit ratio that precludes its use in many digital acquisition, transmission or playback applications. U.S. Patent No. 4.57
In the fourth embodiment described in No. 1.635, ■△XI 1/750 of Digital Equipment Co., Ltd.
It should be noted that a computer was used to execute the image extortion program according to the algorithm disclosed in the patent.

One of the obstacles in implementing the algorithm represented by equation (2,2) in hardware is due to the difficulties in computation-L that arise in connection with taking an intermediate level or intermediate 1tiD into 8 windows. be. The median value or average value of this window is
Multiply the center pixel and surrounding pixels by 1, sum the results, and divide by the number of pixels in the window, e.g. 3×3
In the case of a window, it consists of the result obtained by dividing by 9. Therefore, in the case of 3x3, where the corners are the values of pixels at various nine positions of the 3x3 window.

The main drawback when implementing the intermediate value equation (3,1) is that 9
and the method of calculating the intermediate In, that is, the average value. Traditional methods of dividing by arbitrary numbers include developing specialized processors to handle the division. This solution is expensive because it requires too much circuitry and too many clock cycles.

The simplest way to calculate intermediate values according to equation (3,1) involves keeping all nine numbers in memory and applying them simultaneously to intermediate registers. Once the sum is calculated, it is transferred to the divider. The disadvantage of this configuration is that all the numbers have to be stored in registers until the sum is obtained, which requires a few more bits for representation and thus increases the complexity of the adder, which requires 9 The disadvantage is that division is still required.

What is another way to determine the intermediate value according to equation (3,1)?
! It is the summation of numbers in several stages. The first stage is to add the eight elements in multiple pairs using four adders, and use the most similar one as a carry.

The output of the first stage is one bit more than the input. The second stage sums four of the remaining five numbers, uses the fifth as a carry, and again increases the output word length by one pit. This continues for four stages of adders, and the final word length is 4 pits longer than the largest input word. The output of the fourth stage is passed back to the processor and divided by nine. This pipe line method is fast in that once the first stage finishes processing the input and passes the result to the second stage, it is free to process the next window immediately. The disadvantage of this is that eight adders are required, the complexity of the adders increases at each stage to handle progressively larger word lengths, and you still face the division-by-9 problem. be.

Another obstacle to hardware implementation of the algorithm expressed by equation (2,2) is that the standard deviation of the pixel values is difficult to calculate for an image window of arbitrary size used in determining the coefficient k in the equation. That's true.

The standard deviation of the image window provides a measure of the extent of the image window relative to the median value of the window. In the image processing area, this is also a measurement value of high frequency components. U.S. Patent No. 4,
In the case of masking that is not sharp as in No. 571,635, the amplification degree (coefficient k) of the pixel given the standard i deviation and its size is adjusted &IJllI.

Currently, there are two solutions to this level of real-time computing problem. The first solution requires expensive computer programming to perform the required calculations. The second solution requires the manufacture of a dedicated circuit board configured to perform the calculations. By cascading multiple dividers followed by several stages of multipliers and accumulators and one block of square roots, the calculation of the standard deviation can be compressed onto a few circuit boards. This solution is also expensive to implement and integrate with the unsharp masking algorithm.

[Summary of the Invention 1] The present invention is based on an algorithm of the general form expressed by Equation (2,2), with modifications that eliminate the aforementioned obstacles, and which is far superior to the implementation according to Equation (2,2>). To provide a non-sharp masking enhancement filter which is simple and inexpensive.

The present invention has a large number of pixels, and a central pixel (D
) using a scanning or moving window. In this case, for a given center pixel, 2” surrounding the center pixel of the window without using the center pixel
(N is 1 or more) The average value of the pixels is obtained. Such an average value is called a circular intermediate value m. A window with an odd number of matrices has a center pixel. Each pixel value of the digital image for which a window can be established is used as a center pixel in scanning the image with the window. This is, for example, 3
In the case of a ×3 window, when all pixels surrounding the center pixel are used, the annular intermediate value means that a division by 8 is required instead of 9.

This makes the required circuitry very simple. 2
Since each of the N pixels is processed in the same way to produce a circular intermediate value, using the 2N pixels surrounding the center pixel means that no weighting is applied in the process of determining the circular intermediate 11m. The present invention further improves the center pixel (
A window is used to calculate the sum of differences between Dc) and a selected number of pixels around it (D, ). In the case of a 3x3 window, using all 8 pixels surrounding the center pixel involves calculating 8 summed differences. Such a calculated value is the central deviation σ. It is called. In the case of a 3×3 window, σ −Σ l D, −Dc
l: N-8(4,1) This central deviation calculation is used in place of the standard deviation calculation adopted in U.S. Pat. Adjust. Using the calculation of the central deviation instead of the standard deviation greatly reduces the required circuit complexity and improves the results obtained when using the standard deviation in determining the value of k in equation (2,2>). There is no substantial reduction in image quality compared to what is available.

A further simplification of the technique disclosed in US Pat. No. 4,571.635 is also provided by the present invention. U.S. Patent No. 4.571.635 calculates the standard deviation using Funpyuta! The coefficient k was obtained based on 1IIIl. The present invention eliminates the need for a computer and includes a memory with a look up table.

This look up table calculates the central deviation σ.

and a gain control input C which is used to establish the memory location of the coefficient or modified gain value. Circuitry is also provided for determining an enhanced or modified value (D) for the center pixel of the scan window according to the following equation:

DC'-[f((7,'Cg)X(Dc--ma)]+
ma (5, 1) where n is an integer greater than or equal to 0, and follows the form of equation (2, 2) of the non-sharp masking emphasis filter. In more general form, the formula is not limited to obtaining the modified gain from the center pixel (D, ) and pixels selected from the pixel window, but rather by obtaining an enhanced or modified value (D) for the center pixel. It can be determined by the following formula.

D − [QX (D, −m,)] + ma annular median value m 1 and central deviation σ. Since using L
It can be realized using SI (large scale integration) circuit technology.

Furthermore, features of the invention provide for the determination of the local background available when it is desired to print in a two-level manner, i.e. with or without marks, using pixel enhancement or modification data signals; The method is to use a circular intermediate 1 [ma]. This requires scaling the enhanced data signal into one bit according to a threshold. Once the local background is predicted, a threshold value to be used is predicted and compared with the value of the enhanced pixel signal to determine whether a mark should be generated.

If the emphasized pixel value is determined to be greater than the threshold, a mark is generated when a positive image is generated using this pixel value, or no mark is generated when a negative image is desired. The emphasized or modified pixel value is determined by using the average of the circular intermediate values for the modified pixel and a window centered on the pixel, which is a fixed number of consecutive pixels before and after the modified pixel. A local background prediction is performed for each pixel. In other solutions, before and after the pixel in question where windows do not overlap,
The average value of the circular intermediate values in a fixed number of windows is used. Both solutions use circular intermediate values that are sufficiently close and far enough from the pixel in question to obtain valid predictions about the local background. The threshold value used is obtained from a look up table located in memory. The background value is used in conjunction with the contrast control value to select the threshold value to use from a look up table. This threshold value and the value of the emphasized pixel are input into a comparator to determine whether the emphasized pixel exceeds the threshold value. Whatever type of device or devices may be used to guide the output of the comparator to generate an image or store it in memory for later use. In the case of a positive image, the enhanced pixel value is used to generate a mark by the imaging device or devices when the pixel value exceeds the threshold; When not in use, avoid generating marks.

Novel features and advantages of the present invention will become apparent to those skilled in the art upon consideration of the following detailed description in conjunction with the accompanying drawings.

DETAILED DESCRIPTION OF THE EMBODIMENTS Referring to FIG. 1, a simplified block diagram of a digitized image enhancement circuit using a moving pixel scanning window is shown. The 2' (where N is greater than or equal to 1) pixels surrounding the center pixel are used to obtain 111IDC' enhanced or modified by 1o for each center pixel value (D) of the scanning window. The enhanced or modified center pixel value Dc' is obtained according to the following equation.

DC' = [f (a., C,) x (Do--ma month 10 m.

(6,1) However, f<a, Cg) is the window and hicane tiIjl
a Central deviation σ for input C. at￥: is,
ri-ma is the annular intermediate value of the scanning window. Each enhanced pixel value (D') can be used to generate an image mark, the collection value determining the density of the image mark. For purposes of illustration, a 3x3 pixel scanning window is used in the configuration shown in FIG.

The circuit used to generate the highlighted pixel 11 (D) consists of shift registers 2 and 14, a window shift register,
Register 6, circular intermediate circuit section 18, center! Difference circuit section 2
0, Modified gain (“(C0°C0)) look up
It includes a table memory 22 and a pixel modification circuit section 24.

The image pixel value of each pixel, for example in digital form 8 pixel, is applied to the input 6 of the window 16 and to the shift register 2, eg in series. The shift register 2 has the capacity to store one line of image in order to do one line of IF. Such delays can also be achieved using addressable random access memory. The output of shift register 12 is fed to input 8 of window 16 and to shift register 4, which also provides a one line delay. What is the output of shift register 4? 1i16 input 13
supplied to Referring to window 16, as M16 is clocked, the pixel value applied to input 13 increases! iD1, then shifted to position D, and then shifted to position D3.

Similarly, the pixel value supplied to input 8 of window 16 is at position D
It then appears at position D, and then at position ffD5. Place 1fD6, D7 and Place I! Similarly, for D8, the image collection value inputted from input 6 is taken in and shifted. Thus, each clock signal results in a new set of pixel values forming a window. Each position D1 to D8 supplies an output to the annular intermediate circuit ff118 and the center deviation circuit section 20. The center image collection position D of the window 16 is the center 1ilj element W1
(Dc) is supplied to the pixel modification circuit section 24 and center deviation circuit section 20.

The annular center value (m 2 ) is calculated by 116 annular intermediate circuit sections 18 provided for each center pixel (Dc) and is supplied via an output 53 to the pixel modification circuit section 24 . This circular intermediate value is the center image! This is the average value of the pixels in the two figures of the window 16 surrounding the given center pixel value without using Ia. However, N is 1 or more. In the configuration shown in FIG. 1, a total of 8 pixels surrounding the center pixel of the window 16 are used, so in the annular intermediate 1ifi ill i, after obtaining the sum of 8 pixels 1inD1~D and D5~D8, a v1 body of 8 is used. is necessary. Other ways in which the window of the invention can be easily implemented include adding pairs of pixel values surrounding a center pixel, dividing the sum of these pairs by two, and so on until a single sum is obtained. This includes repeating and dividing by 2 again. The result of division by 2 does not require the use of the lower bits of the binary output (J, so each division by 2 is easily accomplished. A small 61 difference is introduced, but this is problematic. Similarly, groups of four pixel values surrounding the center pixel to be used can be combined with the sum divided by four without using the least significant bit and the next least significant bit in the output bits. An example of division by 2 of the annular intermediate circuit section 18 is shown in FIG. 12 for a 3.times.3 window.

In this case, the four adders 25 to 28 are used to phase the pixels +a D ~1)8 of the 3×3 window 16 using a plurality of pairs. Each input and output of the adder is shown by a tree of lines, but since one pixel is represented by 8 bits, each input line represents 8 input lines and the output line represents 9 lines. Therefore, when the outputs of adders 25 and 26 are applied to adder 29, and when the outputs of adders 27 and 28 are applied to adder 30, the pixel Wi and D2 are added by adder 25, pixel values D2 and D4 are added by adder 26, pixel [D and 1]8 are added by addition 328, and division by 2 is added ti! Execute based on outputs from 25 to 28. Similarly, v1 calculation by 2 is addition f$i! 29
and 30, and the result is applied to the adder 31. A division by 2 is also carried out in the adder 31 and the desired annular intermediate value m 1fiWi of the window is obtained from the output 53 . m2
As can be seen, all adders 25-31 are identical, minimizing size and complexity.

Furthermore, since all pixel values are processed in the same way, no pixel values are disproportionately weighted.

A selected number of pixel values surrounding the center pixel of the pixel window are input into a center deviation calculation in center deviation circuitry 20.

As previously mentioned, all pixel values (D through D8 and DC) in the 3.times.3 window 16 of FIG. The center deviation circuit unit 20 calculates the center deviation (σc
′) is determined. The center deviation (σ0) is the absolute first sum of the differences between the center pixel and the 1iij pixels around the selected window. For example, in the case of a 3 x 3 window such as 3 x 3 window 16,
Take the center pixel and use all pixels including 111. This involves determining eight differences that are summed. Expressed in the form of an equation, σ = Σ l D, -Do I : N = 8C, , + (7, 1) Figure 3 shows that one arm of the center Q difference (σC) shown in equation (1, 1) A central deviation circuit section 2o is shown for performing eight binary subtracters 3.
Center pixel (1) using 2-39. ) and 3x3 window 16
Each peripheral pixel 1a (D.

to D8). As in the circuit of FIG. 2, the multiple inputs of the binary mouthpieces 32-39 are shown by a single line, but one line is used for each of the 8 bits for the pixel values. As shown, each binary payment machine 32 to 39
In this case, one person inputs the center pixel (D.), and in binary subtraction 1i332-39, pixels II (01-D8) are also input, respectively. Each absolute difference is a binary subtractor 32-39
The sum of such differences is outputted from a large number of adders 40 to 40.
46. In particular, the adder 40 adds the outputs of the binary subtraction δ32 and 33, and the adder 41 adds the outputs of the binary subtraction δ32 and δ33.
The outputs of subtracters 4 and 35 are added together, and adder 43 adds the outputs of subtracters 38 and 39. The output lines of each filler 40 to 43 represent nine output lines. The sum output from adders 40 and 41 is input to adder 44, and the sum output from weights I2 and 43 is input to adder 45. The sums output from the 10 output lines of adders 44 and 45 are input to adder 446 and multiplied by 11. The output 47 of each tiJ 46 requires 11 tree output lines'
With C, it becomes a center deviation (σ.) signal.

Center deviation circuit i! Output I determined by i20! The center deviation (σ) of 47 is used when determining the amount of sieving, that is, the emphasis positive, to be input to the center pixel Wi (D.). The center deviation, along with the gain control input 50, is used as the address of a look-up table in the memory 22 that stores the value of the previously ffL I gain υItIl. Editing the look-up table takes into account the absolute difference numbers obtained by calculating the central deviation. If not taken into account, the center deviation determined by the center deviation circuit section 20 is divided by the above number to obtain an intermediate value of the center deviation. With a modification circuit of the type shown in FIG. 3, it is easy to perform such a division to obtain an intermediate value of the center deviation to obtain the output of the adder 46. This output corresponds to the above III! It has only output lines larger than the output line with the most significant bit of the absolute difference number obtained by the Iij circuit. The median value of central deviation is 17
Other possible modifications include addition 34 in Figure 3.
The purpose is to omit the use of the output line for the lower bits that outputs 1 to 46, which in the case of FIG. 3 is output from the adder 46. This output is equal to the sum of the differences obtained from subtractors 32-39 divided by the number of subtractors, ie, eight. When not performing the division to determine the central deviation, f
The resolution of (σc′C0) increases.

The greater the absolute value of the gain control, the greater the enhancement of the image limbus. The gain υ control obtained from the look-up table is input to the pixel modification circuit section 24. The pixel modification circuit unit 24 calculates a modified or emphasized center pixel value (Dc) according to equation (6, 1>).

The pixel modification circuitry 24 is shown in more detail in FIG. 4 and includes a subtracter 52. The subtractor 52 receives the annular medium fffl value (ma) via 53 from the center pixel 1 via 54.
It is input with +11 (D,), and 11 heart pictures! +1
The annular median value (ma) is drawn from 11(D). The multiplication circuit 55 is connected to the look-up table memory 22
57 gain 11111 input value f(σc
'C0), the output of the subtracter 52 (including the sign bit) is input, and the gain control input value (f(σc'c
, >> is multiplied by the output (Dc-ma) from the subtracter 52 to obtain the product. This product, together with the annular intermediate value ma provided via 53, is fed via 58 to an adder 59 which outputs the emphasized or modified center pixel value at output 10.
get. The enhanced or modified pixel values delivered to the output 10 of the pixel modification circuitry 24 are useful in creating images. In this case, the process is performed using 8 pits for each pixel.
An image is created that produces a gray level image defined by the level signal.

An image can be represented by a binary system using digitally determined pixels, that is, by the presence or absence of a mark. This means that the pixel value obtained at the output 10 of the pixel modification circuit 24 must be made into one bit according to the threshold value.

This can be done using a threshold, the selection of which is based on local background measurements. In that case, the threshold value and the emphasized center pixel II (Dc)
Compare between. When the emphasized center pixel value is greater than or equal to a selected threshold value, it provides an output that can be used to generate a mark based on the median value of the image. When the emphasized center pixel value is smaller than the selected threshold value, output is performed without generating a mark based on the intermediate value of the image. By using the annular intermediate value determined by the annular intermediate circuit unit 18 for the center pixel value of the pixel window, a useful measurement of the local background can be made. In order for a measurement of the background of a given enhanced pixel to be useful, the annular intermediate values used must be sufficiently close to, and sufficiently far apart from, the pixel in question.

Such a configuration is schematically shown in FIG. 5. In FIG. 5, 3×3N is used. The window 60 has a central pixel 48 that determines the emphasis value and compares it to a threshold, ! I61 is centered at a pixel five pixels ahead from the center pixel of the window 60, and window 62 is centered at a pixel five pixels behind the center pixel of the window 60. The average value of the annular median values determined for each of these three windows can be used as a background measurement value for the 160 center pixels.

FIG. 6 shows another configuration that can easily implement a circuit using 8-window annular intermediate values. Finding Enhancements or Modifications and Comparing to Thresholds In the case of pixel 63, the annular intermediate values of the four 3x3 non-overlapping windows 64-67 ahead of pixel 63 are It is used together with the annular intermediate value of the compensation windows 68-71. The average value of the annular intermediate values of the eight non-overlapping windows 64-71 is determined and used as the background value for the pixel 63.

FIG. 7 shows a circuit implementing the configuration described in connection with FIG. 6. As is clear from FIG. 6, in the case of 8 annular intermediate values in the windows 64-71 available for averaging, 24 series connected shift registers 172-195
Use. Four framers 96-99 each having two inputs for an 8-bit signal are used to receive 1164-71 circular intermediate values. The input 68a of addition $9'a96 is the window 68
receives the circular intermediate value of . This circular intermediate value is also input to shift register 172. This is adder 99
This occurs when the input 67a of the window 67 is inputting the circular intermediate value of the window 67. The other inputs of the adder 96 are 1. It is connected to the output of the shift register 174 and inputs the circular intermediate value of the window 69. The circular intermediate value of the window 70 is output from the shift register 177, and the addition S! 1i97 is supplied to the input cuff 0a. The other input cuff 1a of the adder 97 is connected to the output of the shift register 180 and receives the annular intermediate value of the window 71. The circular intermediate value of window 64 is output from shift register 186 and provided to input 64a of adder 98. Other input 6 of adder 98
5a is connected to the output of the shift register 189, and inputs the value inside the ring of the window 65. The circular intermediate value of window 66 is output from shift register 192 and added!
N! 1:99 input 66a. The circuit of FIG. 7 also includes adders 100 and 101. Adder 96
The output line of the lowest pit among the nine output lines of
Since it is not connected to the input 102 of 100, adder 9
The sum of the circular intermediate values of 168 and 69 output from 6 is 2
m and is supplied to the adder 100. Similarly, the output of adder 97 is the other input 103 of adder 100.
It is connected to the.

Similarly, the outputs of adders 898 and 99 are connected to inputs 104 and 105 of adder 101, respectively. An adder 106 is also provided, whose inputs are likewise connected to the outputs of adders 100 and 101, respectively. The output of adder 106 lacks the least significant bit of the sum of the two inputs, making its output the average value of the eight circular intermediate values of windows 64-71. The circuit portion that obtains the average of 8 inputs including adders 96 to 106 uses adders 25 to 31 to calculate the average of 8 inputs.
It should be noted that the structure and ms are similar to the part of the circuit of FIG. 2 that obtains the average of the inputs.

Referring to FIG. 1, the circuit that processes the output 10 of the emphasized pixel from the pixel modification circuit section 24 using a threshold value and uses this to binary form to generate an image using the Bell method includes background detection. Circuit portion 1101L includes a memory 111 for a threshold look-up table, a comparator 112, a delay circuit 113, and an inverter 115. Background detection circuitry 110 may take the form of the circuit described in connection with FIG. The background detection circuit section 110 receives the annular intermediate m (ma) from the annular intermediate circuit section 18 and outputs the background value of the emphasized pixel value obtained from the pixel modification circuit section 24 . This background value is used as the address of a threshold look up table in memory 111, which address is selected by table select input 116 to memory 111. The 8-bit threshold value obtained from memory 111 is
It is applied to one input of the 8 bits of comparator 112. The comparator 112 has the delay circuit 11 connected to the other 8-bit input.
Each emphasized pixel value is introduced from the pixel modification circuit y524 via the pixel modification circuit y524. The delay circuit 113 provides a necessary delay to the pixel # value which is supplied to the comparator 112 at the time point 11 when the background value of the pixel value is 11. The amount of delay obtained by the delay circuit 113 depends on the method in which the background value is determined by the background detection circuit 1110 and the method in which Dc' is obtained. Comparator 112 outputs one negative bit when the emphasized pixel value is greater than or equal to the threshold value of the pixel value supplied to comparator 112. Further, the comparator 112 determines that the pixel value obtained by IJm is
When the pixel value supplied to the pixel value is smaller than the threshold value, the other 1 bit is output. The output of comparator 112 is input to inverter 115. The inverter 115 is the comparator 112
The output system is supplied to the inverter 115! ! 11
117, output as is or inverted. When the output of the comparator 112 is output through the inverter 115 when it is greater than or equal to the threshold, it can be used by image processing to generate a mark.

Furthermore, once the output of the comparator 112 is inverted, the image processing uses this to prevent the occurrence of a tone. Similarly, when inverter 115 sends the output obtained from comparator 112 when the emphasized pixel value is less than the threshold value, no mark is generated even if image processing uses this output. However, when the output of comparator 112 is inverted by inverter 115 and used in image processing, a mark is generated.

FIG. 8 shows a modification of the portion shown in FIG. 1 described above.

This transformation is performed based on a halftone method using the enhanced pixel values from the pixel sieving circuit section 24, if necessary.
Bit signals can be created.

This variation includes memory 118 and multiplexer 119.
Includes additions. Memory 118 contains a halftone value look up table.

Selection of the halftone value is performed at the address input 121 of the memory 118 along with the table selection input 120 of the memory 118.
determined by the pixel-based address and line count supplied to the pixel-based address and line count. Multiplexer 119 is memory 1
The output from the computer 11 is input to an input 122, and the output from the computer 118 is input to an input 123. Multiplexer 119
To select one of the input 122 of the memory 111 and the input 123 of the memory 118, select whether to pass it to one of the eight pit inputs of the comparator 112.
ll10124 is input.

Although a 3×3 window 16 is used in FIG. 1, it should be understood that the 3×3 window is for illustrative purposes only. It is only necessary that the windows used have certain properties.

Of course, the larger the window size, the more circuit elements are required to process a larger number of pixel values in succession. The effect obtained by the present invention disclosed herein is that a pixel window for determining the center deviation can be obtained using one center image collection and a plurality of surrounding pixels. The intermediate value and central deviation are numbers that can be expressed as the power of 2. For example, the following pixel values: the first two and last pixel values of the first row, the last pixel value of the second row, the first pixel value of the fourth row, the first and last two pixel values of the fifth row. pixel values in the circular intermediate circuit section 1
When not inputting 8, a 5×5 window can be used. This is the 16 surrounding the center pixel using with a given center pixel, without using the 8 outer pixel values.
(24) means calculating a circular intermediate value of pixel values. A 7x7 pixel window can also be used when calculating the annular intermediate value without using the 4 pixel values at the corners of the 6 window. When 16 pixels were not used, the 32 (25) pixel values surrounding the center pixel were used to create a circular intermediate value of 81
Calculate. When the same pixel window of the 9x9 window is not used, the 9x9 window is appropriate because the 64 (26) pixel values surrounding the center pixel are used to calculate the annular intermediate value. 1
A 3×13 pixel window can also be used. In this case, the pixel values of only positions 6 and 8 in the first and most dominant rows are used, and the pixel values of positions ii! in the first and thirteenth columns are used. Only pixel values of 6 and 8 are used. With such a configuration, a circular intermediate value is calculated using 128 (27> pixel values) surrounding the central pixel value. As another example, if 8 pixels in each color of the window are not used, a 17 x 17 pixel window may be used. Therefore, the circular intermediate value is calculated using 256 (28> pixel values) surrounding the center pixel.Referring to the upper left edge of such a window, the 8 pixel values that are not used for such corners are in the first row. The first four pixels, the first two pixels of the second row, and the third
Preferably, it is the first pixel in the row and the fourth row. The eight unused pixels in the other three corners are selected to obtain the same type of configuration for those corners. A suitable pixel window need not have an IDl-default matrix. For example, Sokari 4
When the two pixels at the corners are not used for the needle bar of the annular intermediate value, a 5×7 pixel window is appropriate, so the annular intermediate value is calculated using 32 (5) pixel values surrounding the center pixel. The above example shows how other larger pixel windows can be used if desired.

When calculating the median value of the center deviation by the center deviation circuit section 20, all pixels covered by the pixel window are available, but only a selected number of available pixel values need to be used. The selected number is taken into account when editing the i-gain lookup table. For example, the selected pixel value is drawn on the display of the pixel window and is contained within a circle of the selected radius, and only the center pixel as the center can give a more accurate center deviation measurement for a particular pixel window. good.

#3El to be determined by the center deviation circuit unit 20 and added to the center pixel value ([), ) using the center deviation (σC)
iI, i.e., the strong Jffi, is incremented by -4. The following explanation indicates that the compilation of the ``qualified gain look-up table'' in memory 22 takes into account the number of absolute differences obtained by the central deviation calculation. Therefore, the division performed by the center deviation circuit 20 is performed by it when editing the memory 22.
It is exposed and exposed. This divides the center deviation circuit part 2
Since there is no part that is truncated or lost in the center deviation circuit section 2o as in the case where the modification gain is determined in the memory 22, the resolution is increased as a result. If such a stubble is not taken into account when editing the look-up table of the method 22, and the pixel modulation following the look-up table editing of the modified gain is performed by the 1-rule circuit. The resolution is improved when the look up table in memory 22 is edited to account for the calculation. The division performed by the pixel modification circuit is
Preferably, the number is such that it facilitates this division, for example by 2. However, it is an integer on nuolX. Preferably, "n' is equal to 3 or 4. Division by pixel modification circuitry requires changes to the circuits shown in FIGS. 1 and 4. The required changes include memory 22 and The first circuit used instead of the pixel modification circuit section 24
This is shown in circuits N22' and 24' of FIG. Circuit portion 22' differs from memory 22 in that the look-up table values are different. Therefore, the modification gain of the circuit section 22' is the modification gain f(σc,
is denoted by f' (σc,C,) rather than C,). The circuit section 24' in FIG. 10 is similar to the pixel fIJ111 circuit Fi in FIG. 1 in that its output formula indicates division by 2.
l 24 and violet, therefore.

Referring to FIG. 10, the multiplication circuit 55' has the same function as tf (a6.cg>x(D, -m,)
instead of multiplication of f'(σc'.

It differs in that it represents the multiplication of Cg)X (D,, ~ma). Furthermore, the output 58 of the multiplication circuit 55' is input to the division circuit 126. The result obtained from multiplication circuit 55' is divided by 2. However, n is an integer of 0 or more, preferably 3 or 4.

The addition t7159' in FIG. 10 differs from the adder 59 in FIG. 1 in that it receives the output 127 of the division circuit 126 as an input instead of receiving the input from the multiplication circuit in FIG.

The output of adder 59', which also receives the annular intermediate value (ma) via 1153, is represented by the equation shown in addition i559'. It is a clay pot. This formula is shown in block 2 of Figure 9.
This corresponds to the equation shown by 4.

The annular intermediate fri(ma) of the central bithread value (D) is:
It should also be understood that the application of such pixel modification is not limited to the particular method of obtaining gain modification from the center pixel value (Dc) of the pixel window and the selected pixel.

Therefore, when we give the general symbol q to the gain modification,
Modified center pixel till using circular median value (ma)
The formula (D) can be expressed as follows.

DC=[ax(D--ma)]+ma The above specific descriptions are for illustrative purposes only and may be modified to a considerable extent without departing from the new suggestions disclosed herein. be. Accordingly, the scope of the invention is intended to be limited only by what is set forth in the claims, and is to be construed in accordance with any view compatible with this patent.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a pixel enhancement device according to an actual example of the present invention, FIG. 2 is a schematic circuit diagram that can be used for the annular intermediate value circuit of FIG. 1, and FIG. 3 is a center deviation diagram of FIG. 1. A schematic circuit diagram that can be used for the circuit section, Figure 4 shows the m4 circuits that can be used for the image correction circuit section in Figure 1, and Figure 5 shows the circuit diagram for determining the background value of the center pixel of the window. A diagram of available pixel windows. Figure 6 is a diagram showing other uses of the pixel window to determine the background value of the center pixel of the window. Figure 7 is a diagram of the circuit used as the background detection circuit in Figure 1. FIG. 8 is a diagram in which a background value is determined using a pixel window not shown in FIG. 6; FIG.
Figure 9 is a diagram showing the modified part of the device in Figure 1 to increase resolution 1y; Figure 10 is a circuit in block diagram form that is used for the pixel modification circuit in Figure 9. This is a diagram. 6...Input, 12.14...Shift register, 16...
...Window shift register (3X3*), 18...
...Annular intermediate circuit section, 20...Center deviation circuit section, 22.22', 24.24'...Circuit section, 47...
...output, 110...background detection circuit section, 111.1
18...Memory, 112...Comparator.

Claims (15)

[Claims] (1) A multilevel signal of a digitized image,
In a digitized image enhancement device for modifying the supplied pixels, a pixel window (16) moving from the pixels supplied to the device;
and the center pixel (D_c) as the pixel window,
and means (6,
12, 14); means (18) for receiving said 2^N pixels surrounding said center pixel and obtaining an annular intermediate value (m_a) of said 2^N pixels; said center pixel (D_c) and said center pixel; Receiving a selected pixel from said pixel window (16) surrounding the pixel and providing a modification gain (g) according to the following formula D_c′=[g×(
and means (20, 22, 24) for supplying a center pixel (D_c') modified according to D_c-m_a)]+m_a. 2. An apparatus for enhancing a digitized image according to claim 1, wherein said last-mentioned means receives said center pixel and a selected pixel surrounding said center pixel from said pixel window, and said center deviation of said center pixel. Means (2) for obtaining output (47) based on (σ_c′)
0), and means () connected to the output (47) to obtain a modification gain (f(σ_c, C_g)) as the modification gain (g).
22), the center pixel (D_c'), the annular intermediate value (m_a)
and the modification gain (f(σ_c, C_g)), and the following formula Dc'=[f(σ_c'C_g)×(D_c-m_a)
]+m_a. 3. An apparatus for enhancing a digitized image according to claim 1, wherein said last-mentioned means receives said center pixel and selected pixels from said pixel window surrounding said center pixel; means (20) for obtaining an output (47) based on said output (47);
means (22') for obtaining a modification gain (f'(σ_c, C_g)) as the central pixel (D_c'), the annular intermediate value (m_a);
and the modification gain (f'(σ_c, C_g)), and the following formula D'=[
f'(σ_c'C_g)×(D_c-m_a)]+m_
Means (24'
) A digitized image enhancement device characterized by comprising: (4) A device for enhancing a digitized image according to claim 2 or 3, characterized in that the means (22, 22') are connected to the output (47).
) is a memory having a modified gain look-up table. (5) The digitized image enhancement device according to claim 2 or 3, wherein the output (47) based on the central deviation is equal to the central deviation divided by the selected number of pixels. emphasizing device. (6) The digitized image enhancement apparatus according to claim 1, 2 or 3, further comprising: a background detection circuit section which receives the annular intermediate value and obtains an output that is a modified background measurement value of the central pixel; 110), means (111) connected to the output of the background detection circuitry (110) for obtaining a threshold; and receiving the threshold at a first input of the comparator and receiving the modified center pixel at a second input of the comparator that obtains a measurement of ground from the background detection circuitry, and determining whether the second input is greater than the threshold at the first input; or a comparator (112) providing one output signal when equal and a different output signal when the modified center pixel is less than the threshold. Image enhancement device. (7) A device for enhancing a digitized image according to claim 5, characterized in that the means (111) for obtaining the threshold value is a memory having a threshold look-up table. Device. 8. An apparatus for enhancing a digitized image as claimed in claim 1, 2 or 3, further comprising a memory having a halftone look-up table, and receiving at a first input thereof a halftone value from said memory; and receiving at a second input thereof the modified center pixel and providing an output signal when the second input is greater than or equal to the halftone value at the first input; a comparator (112) for providing a different output signal when the center pixel, which is the center pixel, is smaller than the halftone value at the first input. (9) A method of enhancing a digitized image for modifying a pixel that is a multilevel signal of the digitized image, comprising: (a) establishing a pixel window moving from the pixel;
(b) having as the pixel window a center pixel (D_c′) and at least 2^N pixels (N is an integer greater than or equal to 1) surrounding the center pixel; The received annular intermediate value (m_a) of the 2^N pixels
(c) establishing a modification gain (g) for each center pixel; and (d) determining the annular intermediate value (m_a), the center pixel (D_
c') and the modification gain (f(σ_c, C_g)) to modify the center pixel (D_c') according to the following formula into the following formula D_c'=[g×(D_c-m_a)]+m_a. A method for enhancing a digitized image, comprising the steps of: (10) The method of enhancing a digitized image according to claim 9, further comprising modifying a pixel that is a multi-level signal of the digitized image, further comprising determining a center deviation of each center pixel established using selected pixels surrounding the center pixel. The step of obtaining each of the modified gains (g), when used by the processing of the final step, satisfies the following equation D_c′=[f(σ_c, C_g)×(D_c−m_a
)]+m_a, the center pixel (D_c'
) A method for enhancing a digitized image, characterized in that a modification gain (f(σ_c, C_g)) of each center pixel is obtained based on a center deviation such that (11) The method of enhancing a digitized image according to claim 9, wherein the pixel of the digitized image is modified as a multi-level signal, comprising: determining a center deviation of each center pixel established using selected pixels surrounding the center pixel; The step of obtaining the modified gain (g) of each center pixel, when used by the final step processing, yields the following equation:
a)] + m_a/2^n (where n is an integer greater than or equal to 0) ,
A method for enhancing a digitized image, characterized in that C_g)) is obtained. (12) A method for enhancing a digitized image according to claim 10 or 11, wherein the step of establishing the modification gain comprises modifying a pixel that is a multi-level signal of the digitized image.
A method for enhancing a digitized image, comprising a memory having a look-up table, using said center deviation to address said memory and to establish said modification gain. (13) The method for enhancing a digitized image according to claim 10, 11 or 12, further comprising modifying a pixel that is a multi-level signal of the digitized image, further using a circular intermediate value (m_a) to modify the modified central pixel. obtaining a measurement of a background; establishing a threshold based on the measurement of the background; and comparing the threshold to the modified pixel to determine whether the modified pixel is the providing one signal when the modified pixel is greater than or equal to a threshold, and providing the other signal when the modified pixel is less than the threshold. Method. 14. The method of claim 13, wherein the step of establishing a threshold comprises a memory having a threshold look-up table. . A method for enhancing a digitized image, characterized in that the background measurements are used to address the memory and establish the threshold. (15) The method of enhancing a digitized image according to claim 9, 10, 11 or 12, further comprising a memory having a look-up table of halftone values, for modifying pixels that are multi-level signals of the digitized image. comparing the halftone values obtained from the memory and providing one signal when the modified pixel is greater than or equal to the halftone value; A method for enhancing a digitized image, characterized in that the method comprises the steps of: providing a different signal when the signal is small.